(self)
| 116 | assert fd_config.scheduler_config.max_num_batched_tokens == 8192 |
| 117 | |
| 118 | def test_fdconfig_init_cache(self): |
| 119 | parallel_config = ParallelConfig({}) |
| 120 | graph_opt_config = GraphOptimizationConfig({}) |
| 121 | cache_config = CacheConfig({}) |
| 122 | cache_config.cache_transfer_protocol = "rdma,ipc" |
| 123 | cache_config.pd_comm_port = "2334" |
| 124 | load_config = LoadConfig({}) |
| 125 | scheduler_config = SchedulerConfig({}) |
| 126 | scheduler_config.splitwise_role = "prefill" |
| 127 | model_config: Mock = Mock() |
| 128 | model_config.max_model_len = 512 |
| 129 | model_config.architectures = ["test_model"] |
| 130 | model_config.mm_max_tokens_per_item = None |
| 131 | |
| 132 | fd_config = FDConfig( |
| 133 | parallel_config=parallel_config, |
| 134 | graph_opt_config=graph_opt_config, |
| 135 | cache_config=cache_config, |
| 136 | load_config=load_config, |
| 137 | scheduler_config=scheduler_config, |
| 138 | model_config=model_config, |
| 139 | test_mode=True, |
| 140 | ) |
| 141 | fd_config.init_cache_info() |
| 142 | assert fd_config.register_info is not None |
| 143 | |
| 144 | def test_fdconfig_postprocess_ports(self): |
| 145 | data_parallel_size = 4 |
nothing calls this directly
no test coverage detected